Ollama APIs: REST and OpenAI-Compatible

The Local HTTP API

Ollama listens on port 11434 by default. Any language can send a generate request over HTTP.

curl http://localhost:11434/api/generate 
  -d '{"model": "llama3.2", "prompt": "Explain recursion in one sentence", "stream": false}'

OpenAI-Compatible Endpoint

Point existing OpenAI clients at Ollama by swapping the base URL. Code and tools that speak OpenAI work unchanged.

from openai import OpenAI

client = OpenAI(base_url="http://localhost:11434/v1")

reply = client.chat.completions.create(
    model="llama3.2",
    messages=[{"role": "user", "content": "Say hello in five languages"}],
)
print(reply.choices[0].message.content)

Official Python Client

pip install ollama

import ollama
resp = ollama.generate(model="llama3.2", prompt="Write a haiku about waves")
print(resp["response"])

Streaming

Most endpoints stream tokens by default. Set stream to false when you only need the final answer, and keep it true for chat-style interfaces.

Key Points

• The API lives at localhost:11434.
• /v1 endpoints speak the OpenAI format.
• The ollama Python client wraps the REST API.
• Disable streaming for one-shot, scripted calls.